Novel Models For Prediction And Diagnosis Of Bacterial Infection In Hepatitis B Virus-related Acute-on-Chronic Liver Failure Based On Inflammatory Factors

BackgroudPatients with acute-on-chronic liver failure(ACLF)are susceptible to bacterial infection owing to intestinal bacterial translocation,immune dysfunction,genetic susceptibility,etc.Bacterial infection is not only a common complication of ACLF,but also an independent risk factor for the short-term prognosis of ACLF.Early diagnosis and treatment of bacterial infection in ACLF can significantly improve survival rate of patients.Therefore,establishing models for early prediction and diaganosis of bacterial infection in ACLF has great value of research and significance of clinical practice.Immune injury plays an important role in the development of ACLF.Recent studies have shown that pro-inflammatory factors such as interleukin-6(IL-6),interleukin-8(IL-8)and tumor necrosis factor ?(TNF?)and anti-inflammatory factors such as interleukin-4(IL-4),interleukin-10(IL-10)produced by activated immune cells can lead to systemic inflammatory response syndrome(SIRS),compensated anti-inflammatory response syndrome and mixed antagonist response syndrome,which is one of the pathophysiological basis of bacterial infection development.Patients with persistent SIRS are susceptible to sepsis,resulting in multiple organ failure,which is an important pathological basis of death of ACLF patients.At the same time,in the early stage of bacterial infection development,pathogenic microorganisms can stimulate macrophages,natural killer cells and other cells to release IL-6,IL-8,IL-10 and other inflammatory factors,which play a critical role in the early control of bacterial infection.Therefore,inflammatory factors can be used as potential pivotal biomarkers to reflect the development,progression and prognosis of ACLF with bacterial infection.Currently,various prognostic models of ACLF established in domestic and foreign studies are mainly used to predict the outcome of ACLF itself,assisting clinicians to make decisions regarding medical treatment or liver transplantation.The model for end-stage liver disease(MELD)and MELD-sodium scores are widely used for evaluating the severity of critically ill liver disease.Recently,the chronic liver failure consortium-organ failure(CLIF-C OF),CLIF-C ACLF,Tongji prognostic predictior model(TPPM)and Asian Pacific Association for the Study of the Liver-ACLF Research Consortium(AARC)scores were shown to be superior to the MELD and MELD-sodium scores in predicting the outcome of ACLF.However,there is a lack of a specific prognostic model for HB V-ACLF patients with bacterial infection,with a potential life-threatening clinical consequence.Although some studies have shown that prophylactic antibiotics can reduce the incidence of bacterial infection to improve the prognosis of patients with a high risk of bacterial infection development.The prophylactic antibiotics strategy for bacterial infection in ACLF is undefined,there is still a lack of study on predicting the risk of bacterial infection development in ACLF,which limits clinicians' early identification of the risk of bacterial infection occurrence and the formulation of preventive treatment strategy.Previous studies have shown that some inflammatory factors play an important role for the bacterial infection development.Johanie et al.demonstrated that IL-6 produced by activated macrophages in the gut can increase intestinal permeability,which leads to bacterial translocation that is prone to bacterial infection development.Furthermore,it is reported that the elevated levels of IL-10 can inhibit the immune cells to clear the bacteria when the body is suffering from acute bacterial invasion,leading to the bacterial infection development.Therefore,inflammatory factors can predict the occurrence of bacterial infection in ACLF.At present,the diagnosis of bacterial infection in ACLF is mainly based on clinical symptoms and signs,laboratory tests and imaging examinations.But there are still some unsolved problems in the above diagnostic criteria.Firstly,the diagnostic efficiency of laboratory parameters for bacterial infection in ACLF is still needed to be improved.Although C-reactive protein(CRP)and procalcitonin(PCT)have certain diagnostic value of bacterial infection in liver cirrhosis,their diagnostic performance of bacterial infection in ACLF is not good.Therefore,a better biomarker or diagnostic system is needed to optimize the diagnosis of bacterial infection in ACLF.Some clinical retrospective studies have confirmed that some inflammatory factors can be used as biomarkers for the diagnosis of bacterial infection,IL-6 can be used for the detection of bacterial infection in patients with liver cirrhosis,with a sensitivity and specificity of 97.5%and 80.6%,respectively.Secondly,some ACLF patients with bacterial infection have mild clinical symptoms or lack of specific clinical manifestations,which are easily ignored by clinicians.Although SIRS can be used to indicate the occurrence of bacterial infection,ACLF itself can lead to the SIRS.How to distinguish the SIRS caused by ACLF or ACLF with bacterial infection has become one of the critical problems in the clinical studies of ACLF.Based on the above research status,we carried out the three clinical studies:exprsssion characteristics of inflammatory factors and establishment of a prognostic model in hepatitis B virus-related ACLF(HB V-ACLF)patients with bacterial infection,establishment of a predictive model for bacterial infection in HB V-ACLF based on inflammatory factors,establishment of a diagnostic model for bacterial infection in HB V-ACLF based on inflammatory factors,which are expected to provide new clinical models for the prediction and diagnosis of bacterial infection in HB V-ACLF.Part I Expression Characteristics of Inflammatory Factors and Establishment of a Prognostic Model in HBV-ACLF Patients with Bacterial InfectionObjectiveTo explore the expression characteristics of inflammatory factors in HBV-ACLF patients with bacterial infection,and to develop and verify a prognostic model for predicting 30-day mortality of HBV-ACLF patients with bacterial infection.MethodsPatients diagnosed with HBV-ACLF at the Tongji Hospital,Tongji Medical College,Huazhong University of Science and Technology from June 2018 to December 2020 were prospectively screened and enrolled for establishing a prospective cohort.Patients diagnosed with HBV-ACLF at the Tongji Hospital,Tongji Medical College,Huazhong University of Science and Technology from January 2016 to May 2018 were retrospectively screened and enrolled for establishing a retrospective cohort.Patients were divided into the no bacterial infection and bacterial infection groups according to the presence of bacterial infection,and patients in the bacterial infection group in the prospective and retrospective cohorts were enrolled into the derivation and validation cohorts to establish and verify the prognostic model of HBV-ACLF patients with bacterial infection,respectively.Data of demographics,medical history,laboratory tests,imaging examinations and survival rate were collected.In the prospective cohort,the clinical characteristics,especially the serum levels of inflammarory factors of patients between the no bacterial infection and bacterial infection groups were compared.The dynamic changes in serum levels of inflammatory factors,30-day and 90-day mortality were recorded.The univariate and multivariate logistic regression analyses were used to determine whether bacterial infection was an independent risk factor for short-term prognosis of HB V-ACLF patients.In the derivation cohort,the univariate and multivariate logistic regression analyses were used to identify the independent predictors for 30-day mortality of HBV-ACLF patients with bacterial infection,then the prognostic model was established,and the corresponding nomogram was drawn.The area under the receiver operating characteristic curve(AUC)was used to evaluate the discrimination performance of the prognostic model,the Hosmer-Lemeshow test combined with calibration curve were used to evaluate the calibration ability of the prognostic model,and the decision curve analysis(DCA)was used to analyze the clinical utility of the prognostic model.ResultsA total of 513 HBV-ACLF patients were enrolled into the prospective cohort,including 289 patients in the no bacterial infection group and 224 patients in the bacterial infection group(enrolled into the derivation cohort).A total of 390 HBV-ACLF patients were enrolled into the retrospective cohort,including 198 patients in the no bacterial infection group and 192 patients in the bacterial infection group(enolled into the validation cohort).In the prospective cohort,the serum levels of PCT,CRP,soluble interleukin-2 receptor(sIL-2R),IL-6,IL-10 and TNF? of patients in the bacterial infection group were significantly higher than those of patients in the no bacterial infection group(P<0.05).The dynamic changes in serum levels of sIL-2R,IL-6,IL-8 and IL-10 of HBV-ACLF patients with bacterial infection were associated with short-term prognosis.The univariate and multivariate logistic regression analyses showed that bacterial infection was an independent risk factor for the 30-day(OR=2.411,95%CI:1.286-4.521,P=0.006)and 90-day(OR=2.477,95%CI:1.377-4.457,P=0.001)mortality of HBV-ACLF patients.In the derivation cohort,the univariate and multivariate logistic regression analyses showed that age(OR=1.073,95%CI:1.035-1.112,P<0.001),total bilirubin(OR=1.006,95%CI:1.003-1.009,P<0.001),international normalized ratio(INR)(OR=1.676,95%CI:1.138-2.648,P=0.009),lactate dehydrogenase(LDH)(OR=1.010,95%CI:1.005-1.015,P<0.001)and sIL-2R(OR=1.001,95%CI:1.000-1.001,P<0.001)were independent predictors for the 30-day mortality of HBV-ACLF patients with bacterial infection.Based on the logistic regression analysis,the five independent predictors were used to constructed a nomogram for predicting the 30-day mortality of HBV-ACLF patients with bacterial infection.In the derivation cohort,the AUC of the nomogram for predicting the 30-day mortality of HBV-ACLF patients with bacterial infection was 0.883(95%CI:0.839-0.927).In the validation cohort,the AUC of the nomogram for predicting the 30-day mortality of HBV-ACLF patients with bacterial infection was 0.852(95%CI:0.791-0.913).In the both cohorts,the AUC of the nomogram for predicting the 30-day mortality of HBV-ACLF patients with bacterial infection was significantly higher compared with other prognostic models(AARC,MELD,MELD-sodium,TPPM,CLIF-C OF and CLIF-C ACLF scores)(P<0.05).Furthermore,the Hosmer-Lemeshow test and calibration curve showed a good calibration ability of the nomogram in the derivation and validation cohorts(P=0.576,P=0.249).The DCA confirmed the clinical utility of the nomogram.ConclusionBacterial infection is an independent risk factor for the short-term prognosis of HBV-ACLF patients,which can lead to the increase of inflammatory factors levels,and the dynamic changes in serum levels of some inflammatory factors are closely correlated to the short-term prognosis of patients.Age,total bilirubin,INR,LDH and sIL-2R were independent predictors for the 30-day mortality of HBV-ACLF patients with bacterial infection.The nomogram constructed by them showed a good discrimination performance,calibration ability and clinical utility.Part II Establishment of a Predictive Model for Bacterial Infection in Hepatitis B Virus-related Acute-on-Chronic Liver Failure Based on Inflammatory FactorsObjectiveTo establish and verify the predictive model of bacterial infection development in HBV-ACLF.MethodsPatients who never developed bacterial infection during hospitalization and developed bacterial infection in the first week of admission in the prospective and retrospective cohorts at the first part were screened and enrolled into the derivation and validation cohorts,respectively.Patients developed bacterial infection in the first week of admission were enrolled in the bacterial infection development group,and patients never developed bacterial infection during hospitalization were enrolled in the no bacterial infection development group.Data of demographics,medical history,laboratory tests,imaging examinations and survival rate were recorded.The incidence of bacterial infection in HBV-ACLF patients after admission was closely observed and followed up.The univariate and multivariate logistic regression analyses were used to identify the independent predictors of bacterial infection development for the establishment of the predictive model,and were constructed correponding nomogram.The AUC was used to evaluated discrimination performance of the nomogram,the Hosmer-Lemeshow test combined with calibration curve were used to evaluated calibration ability of the nomogram,and clinical utility of the nomogram was assessed by the DCA.ResultsA total of 382 HBV-ACLF patients were enrolled into the derivation cohort,and a total of 284 HBV-ACLF patients were enrolled into the validation cohort.In the derivation cohort,the univariate and multivariate logistic regression analyses demonstrated that serum levels of total protein(OR=0.854,95%CI:0.814-0.895,P<0.001),CRP(OR=1.117,95%CI:1.075-1.160,P<0.001)and IL-6(OR=1.025,95%CI:1.006-1.045,P=0.009)were independent predictors of bacterial infection development in HBV-ACLF patients.These three independent predictors were assembled into a nomogram to predict the bacterial infection development.In the derivation cohort,the discriminative accuracy(AUC)of the nomogram was 0.840(95%CI:0.794-0.885).In the validation cohort,the discriminative accuracy(AUC)of the nomogram was 0.794(95%CI:0.738-0.850).The calibration curve and Hosmer-Lemeshow test showed a good calibration ability of the nomogram in the derivation and validation cohorts(P=0.196,P=0.415).DCA confirmed the clinical utility of the nomogram.ConclusionSerum total protein,IL-6 and CRP levels were independent predictors of bacterial infection development in HBV-ACLF patients.The predictive model constructed by them showed a good discrimination performance,calibration ability and clinical utility,which may provide an auxiliary solution for early prevention and treatment of bacterial infection in HBV-ACLF.Part III Establishment of a Diagnostic Model for Bacterial Infection in Hepatitis B Virus-related Acute-on-Chronic Liver Failure Based on Inflammatory FactorsObjectiveTo establish and verify the diagnostic model of bacterial infection in HBV-ACLF.MethodsPatients in the prospective and retrospective cohorts at the first part were enrolled into the derivation and validation groups,respectively.Data of demographics,medical history,laboratory tests,imaging examinations and survival rate were recorded.Patients were divided into the no bacterial infection and bacterial infection groups according to the presence of bacterial infection.The univariate and multivariate logistic regression analyses were used to identify the variables for establishing the diagnostic model of bacterial infection.The AUC was used to evaluated the discrimination performance of the diagnostic model,the calibration curve combined with Hosmer-Lemeshow test were used to evaluated the calibration ability of the diagnostic model,and the clinical utility of the diagnostic model was assessed by the DCA.ResultsA total of 513 HBV-ACLF patients were enrolled into the derivation cohort,including 289 patients in the no bacterial infection group and 224 patients in the bacterial infection group.A total of 390 HBV-ACLF patients were enolled into the validation cohort,including 198 patients in the no bacterial infection group and 192 patients in the bacterial infection group.CRP,PCT,IL-6 and IL-10 were identified by the univariate and multivariate logistic regression analyses to construct the diagnostic model of bacterial infection:Logit(P)=-4.861+0.061×CRP(mg/L)+3.343×PCT(ng/mL)+0.032×IL-6(pg/mL)+0.066×IL-10(pg/mL).In the derivation cohort,the AUC of the diagnostic model for diagnosing bacterial infection in HBV-ACLF was 0.897(95%CI:0.871-0.923).In the validation cohort,the AUC of the diagnostic model for diagnosing bacterial infection in HBV-ACLF was 0.855(95%CI:0.816-0.894).The Hosmer-Lemeshow test and calibration curve showed a good calibration ability of the diagnostic model in the derivation and validation cohorts(P=0.387,P=0.229).The DCA confirmed the clinical utility of the diagnostic model.ConclusionThe diagnostic model of bacterial infection constructed by CRP,PCT,IL-6 and IL-10 showed a good discrimination performance,calibration ability and clinical utility in HBV-ACLF,which can assist clinicians in diagnosing bacterial infection in HBV-ACLF.